Understanding InnoDB X-Lock with REPEATABLE_READ and READ_COMMITED isolation level

52 views Asked by At

I have two mysql(AWS Aurora) db tables:

--------------------------------
|     Table:deparment          |
--------------------------------
--------------------------------
|    id      |  dept_name      |
--------------------------------
|     d1     |       dept1     |
--------------------------------
|     d2     |       dept2     |
--------------------------------
--------------------------------

-------------------------------------------------
|                Table:employee                 |
-------------------------------------------------
-------------------------------------------------
|   id      |   emp_name       |    dept_id     |
-------------------------------------------------
|   e1      |   Emp 1          |    d1          |
-------------------------------------------------
|   e2      |   Emp 2          |    d1          |
-------------------------------------------------
|   e3      |   Emp 3          |    d1          |
-------------------------------------------------
|   e4      |   Emp 4          |    d2          |
-------------------------------------------------

Business logic: Multiple thread can perform create/modify or delete operation on Depatment and Employee table. Any modification on employee table will result updating the employee id

------
Java Code with Spring Transaction:

/**
*   Modify Employee
*/
@Transactional(isolation = REPEATABLE_READ)
public void modifyEmployeeByDeptarment(int deptId, int empId Update upd) {
    Department dept = deptDB.getDeparmentForUpdate(deptId); // returns SELECT * FROM department where id = :deptId FOR UPDATE; Row level Lock on dept
    
    Employee employee = empDB.getEmployee(deptId, employee.id); // returns SELECT * from employee where dept_id = :deptId;
    
    employee.setId(newEmpId);
    addUpdate(employee, upd); // This will update in DB
}

/**
*   Delete department with employees
*/
@Transactional(isolation = REPEATABLE_READ)
public void deleteDepartment(int deptId) {
    Department dept = deptDB.getDeparmentForUpdate(deptId); // returns SELECT * FROM department where id = :deptId FOR UPDATE; Row level Lock on dept

    List<Employee> employees = empDB.getEmployeesByDept(deptId); // returns SELECT * from employee where dept_id = :depltId;

    deleteDepartment(dept); // Deletes dept from DB
    deleteEmployees(employees); // Delete all employee for deptId
}
------

As you can see here, for both the operation I am taking lock on the department row before making any changes to its corresponding employees. But in few cases, It has been observed that employee record was present in our database without its corresponding department.

Can someone please explain this behaviour?

Spring Framework version: 5.2.0.RELEASE Aurora MySQL Version: 5.7.12

* Note: I changed the isolation level to READ_COMMITED, and everything was working as expected.
0

There are 0 answers